The $2.5 Million Lesson
In 2019, a team at the Francis Crick Institute thought they had engineered the perfect CRISPR edit. They wanted to disable a single gene in human embryos to study early development. The target was clear, the guide RNA was precisely designed, and the molecular scissors should have cut exactly where intended. Instead, when they sequenced the results, they found something that made their stomachs drop: CRISPR had deleted enormous chunks of chromosome 4, sometimes removing over 4,000 base pairs in a single swipe.
This wasn’t supposed to happen. The textbook version of CRISPR describes a precise molecular scalpel that makes clean cuts at predetermined locations. But the Crick Institute team had stumbled into one of gene editing’s most uncomfortable truths: our molecular tools are far messier than we’d like to admit. Their failed experiment, published in Cell, revealed that CRISPR sometimes behaves more like a molecular chainsaw than a scalpel, particularly in early embryos where DNA repair mechanisms are still developing.
The researchers didn’t hide their results or bury them in supplementary data. They published everything, including the spectacular failures. This kind of scientific honesty is increasingly valuable as we navigate the gap between CRISPR’s promise and its current limitations.
Off-Target Effects: The Ghost in the Machine
Every CRISPR researcher has a folder of experiments that didn’t work as planned. Sometimes the edits miss their target entirely. Other times, they hit multiple locations across the genome, creating a biological equivalent of friendly fire. In 2018, researchers at Columbia University discovered that their CRISPR edits in mouse retinas had caused over 1,500 unintended mutations scattered across the genome.
The Columbia study was particularly jarring because it contradicted earlier research suggesting CRISPR was highly specific. The team used whole-genome sequencing instead of the more limited targeted analysis that had been standard practice. When they looked everywhere instead of just where they expected to find changes, the landscape looked completely different. The mice appeared healthy despite carrying these mutations, but the implications for human applications were sobering.
These off-target effects aren’t random accidents. They follow predictable patterns based on DNA sequence similarity and chromatin structure. Researchers have developed increasingly sophisticated prediction algorithms to anticipate where CRISPR might cut unintentionally. But predicting and preventing are different challenges entirely, and the prediction tools are still catching up to the complexity of real biological systems.
The Delivery Problem Nobody Talks About
Even when CRISPR works perfectly in a test tube, getting it into the right cells in a living organism remains one of the field’s most persistent challenges. The most publicized CRISPR failures often stem not from the gene editing itself, but from delivery systems that don’t work as intended.
Consider the case of CTX001, a CRISPR therapy for sickle cell disease. The treatment works by editing patients’ bone marrow cells outside the body, then reinfusing them. In clinical trials, some patients showed dramatic improvements, with their bodies producing healthy hemoglobin for the first time in years. But others saw minimal benefits because too few of their edited cells survived the process of extraction, editing, and reintroduction.
The delivery challenge gets even more complex for in vivo editing, where CRISPR must reach target tissues through the bloodstream. Lipid nanoparticles, the current gold standard for delivery, tend to accumulate in the liver regardless of their intended destination. This is why most successful CRISPR therapies so far have targeted either blood disorders (where cells can be edited outside the body) or liver diseases (where the natural biodistribution actually helps).
When Good Edits Go Bad
Sometimes CRISPR works exactly as designed, but the biological consequences are more complex than anticipated. The CCR5 gene is a poster child for this phenomenon after the He Jiankui scandal in 2018, but the underlying science reveals a more complex story about unintended consequences.
CCR5 encodes a protein that HIV uses to enter immune cells. People with natural CCR5 mutations are largely resistant to HIV infection, making it an attractive target for gene editing. But CCR5 isn’t just an HIV receptor. It plays roles in immune function, brain development, and response to other infections. Studies have shown that people with disabled CCR5 genes have increased susceptibility to West Nile virus and more severe outcomes from influenza.
This biological complexity means that even successful edits can create new vulnerabilities. The immune system is particularly challenging because it evolved as an interconnected network. Changing one component often has ripple effects that only appear over time or under specific environmental pressures.
Building Better Tools from Beautiful Failures
The most valuable CRISPR failures are the ones that reveal fundamental limitations in our current approaches. Each spectacular mistake teaches us something new about biology and pushes the field toward more sophisticated tools.
Base editors and prime editors emerged partly from researchers’ frustrations with standard CRISPR’s tendency to create double-strand breaks that cells often repair imprecisely. These newer tools can make single-letter changes to DNA without cutting both strands, dramatically reducing unwanted mutations. They’re not perfect either, but they represent evolution in response to well-documented failure modes.
The field is also developing better ways to measure success and failure. Older studies often used limited sequencing approaches that could miss off-target effects. Now researchers routinely use unbiased genome-wide methods to detect unintended changes. This more comprehensive monitoring doesn’t prevent failures, but it makes them visible and quantifiable.
What I find most fascinating about CRISPR’s development is how openly researchers share their failures. The scientific literature is packed with papers documenting what doesn’t work, often in incredible detail. This transparency speeds up progress because other researchers can learn from mistakes without repeating them. In a field moving as quickly as gene editing, failed experiments published promptly are often more valuable than successful ones published years later.